An AI Artist’s Twitter Feed Is an Art Gallery

#BigGAN generated image by Mario Klingemann

Call it social media performance. Although the artist Mario Klingemann has exhibited his work in prestigious venues such as the Museum of Modern Art in New York and the Centre Pompidou in Paris, he also shares his artistic experiments in informal, “fire and forget” tweets. The images and videos he posted last fall under the hashtag #BigGAN can only be fully understood and appreciated by treating his Twitter feed from the time as a digital exhibition. Over the course of several weeks in November and December, these dispatches — which form a kind of narrative performance — developed an extended metaphor of digital space as a terra incognita, a previously hidden, alien world awaiting cartographic mapping.

Klingemann has used artificial neural networks in his art for years. In 2018 he won the Lumen Prize. Notably, it was the first time the prize went to an artist whose artwork had been generated using artificial intelligence. BigGAN refers to a paper published last September on machine learning techniques that can be used to create strikingly realistic artificial images. A series of joyful, jokey #BigGAN posts followed on Twitter almost immediately, as a community of artists who work with code started to share the results of generating all manner of impossibilites through this technology, including visual puns, surreal clocks, dreamscapes, almost-faces, and the occasional poetic trance. When Klingemann joined in mid-November, he characterized his unique way of coaxing astonishing results from the same algorithms as a form of mapmaking: “All the images are already there,” he tweeted, “I am just finding them.” Finding, in this case, alludes to the ongoing question of how to define originality in relation to the type of generative digital art that Klingemann practices: if the image-making is done by software, and this particular software was developed by Google, then where is the originality? Among the seemingly infinite possible arrangements of pixels on a screen, the art, Klingemann suggests, is in locating a handful that mean anything at all.

#BigGAN generated image by Mario Klingemann

Partly curatorial and partly computational, this mapmaking is as unique to Klingemann as a brushstroke is to a painter. While a newcomer to #BigGAN might struggle for days to find one interesting image among the nonsense, Klingemann’s art showcases his ability to rapidly produce a prodigious body of images, each image more stunning and suggestive than the next. These findings — eerie and ethereal abstractions he posts as either single images or closely related groupings — seem like snapshots from a fable in a language that does not exist. “Optimizing for unrecognizability,” reads one tweet, gesturing at paradox. #BigGAN culminated in a series of breathtaking GIFs and videos by early December, but it was the narrative territory Klingemann covered during this period that ultimately made his tweets into a wholly original undertaking.

#BigGAN generated image by Mario Klingemann

“Found a hideaway in the wastelands of #BigGAN,” one thread began. “I am a bit worried about my food supplies.” Subsequent posts in the thread read like missives from a stranded explorer trying to survive on a baffling, somewhat hostile new planet:

I am in luck. Last night’s sandstorm has brought a flock of Tumbleweasels right into my camp.

Some figures are approaching, but it is hard to make out their shapes with the light playing tricks on me.

I must have lost consciousness. When I woke up the vessel was gone and I can’t detect any traces of its passengers.

But wait. There are some pictures on my recording device I don’t recall having taken.

#BigGAN generated image by Mario Klingemann

The narrative moved apace, in tweet after tweet, each one accompanied by images found along Klingemann’s synthetic journeying. During this same period, Klingemann posted casual tweets about his technical process, details of real trips he took, and responses to comments and questions from his followers, moving fluidly from fact to fiction, from the persona of a scrupulous technician to one of a high magician of synthetic surrealism. The pleasure of experiencing this work in real time was heightened by the disjunctive effect of Twitter’s reverse-chronological interface. Because recent posts were parts of multiple, thematically intertwined threads, I would return again and again to the older threads, as each fresh post added a new beat to their narrative arc. My tight re-readings felt like a way to compensate for not having clear endings. Any day now, Klingemann might pick up one of those threads and add another post to the story, as if no time had passed; any day now, I might casually glance at my Twitter feed to see if he has.

#BigGAN generated image by Mario Klingemann